Sparse Dowker nerves (Q2324601)

The authors propose sparse versions of Dowker complexes and nerve complexes for use in the computation of persistent homology from point cloud data. As initial motivation, their goal was to generalize an interleaving guarantee for the sparse Čech complex to hold not only for convex metric spaces, but also for arbitrary metric spaces. This goal has been realized with a ``truncated'' sub-complex of the Čech complex. Indeed, let \(P\) be an ordered finite subset of a metric space \(M\), such that each \(p_i\in P\) is the furthest point in \(M\) from the set of preceding points in \(P\). Whereas the Čech complex of \(P\) is the nerve of balls in \(M\) centered at the points in \(P\), the \textit{sparse Dowker nerve} is an altered version thereof with fewer simplices. Roughly speaking, the sparse Dowker nerve is obtained by truncating balls to not grow beyond a certain radius (depending on the location of the ball's center), and furthermore ball centers do not appear as vertices of simplices whose filtration time is too large (again depending on the location of the ball's center). Theorem 1 in this paper proves that the sparse Dowker nerve of \(P\) is nevertheless multiplicatively interleaved with the Čech complex of \(P\). This result is only a particular instance of a more general framework, which is constructed using relations, filtered Dowker complexes, interleavings, and 2-categories.